Amortized Safe Active Learning for Real-Time Data Acquisition: Pretrained Neural Policies from Simulated Nonparametric Functions
This work addresses the problem of slow decision-making in safe AL for real-time applications, offering a modular solution that can be adapted to time-sensitive tasks, though it is incremental as it builds on existing amortized Bayesian experimental design methods.
The paper tackled the computational inefficiency of safe active learning (AL) for real-time data acquisition by proposing an amortized framework that replaces online Gaussian process updates and constrained optimization with a pretrained neural policy, achieving substantial speed improvements while maintaining safety and learning quality.
Safe active learning (AL) is a sequential scheme for learning unknown systems while respecting safety constraints during data acquisition. Existing methods often rely on Gaussian processes (GPs) to model the task and safety constraints, requiring repeated GP updates and constrained acquisition optimization-incurring in significant computations which are challenging for real-time decision-making. We propose an amortized safe AL framework that replaces expensive online computations with a pretrained neural policy. Inspired by recent advances in amortized Bayesian experimental design, we turn GPs into a pretraining simulator. We train our policy prior to the AL deployment on simulated nonparametric functions, using Fourier feature-based GP sampling and a differentiable, safety-aware acquisition objective. At deployment, our policy selects safe and informative queries via a single forward pass, eliminating the need for GP inference or constrained optimization. This leads to substantial speed improvements while preserving safety and learning quality. Our framework is modular and can be adapted to unconstrained, time-sensitive AL tasks by omitting the safety requirement.